core: improve OsStr(ing) helpers

This adds the `os_str_as_bytes_lossy` function, for when we want infallible
conversion across platforms, and improves the doc comments of similar functions
to be more accurate and better formatted.

src/uucore/src/lib/features/quoting_style.rs

@@ -8,8 +8,6 @@
 use std::char::from_digit;
 use std::ffi::{OsStr, OsString};
 use std::fmt;
-#[cfg(unix)]
-use std::os::unix::ffi::{OsStrExt, OsStringExt};
 
 // These are characters with special meaning in the shell (e.g. bash).
 // The first const contains characters that only have a special meaning when they appear at the beginning of a name.
@@ -462,36 +460,18 @@ fn escape_name_inner(name: &[u8], style: &QuotingStyle, dirname: bool) -> Vec<u8
 
 /// Escape a filename with respect to the given style.
 pub fn escape_name(name: &OsStr, style: &QuotingStyle) -> OsString {
-    #[cfg(unix)]
+    let name = crate::os_str_as_bytes_lossy(name);
-    {
+    crate::os_string_from_vec(escape_name_inner(&name, style, false))
-        let name = name.as_bytes();
+        .expect("all byte sequences should be valid for platform, or already replaced in name")
-        OsStringExt::from_vec(escape_name_inner(name, style, false))
-    }
-    #[cfg(not(unix))]
-    {
-        let name = name.to_string_lossy();
-        String::from_utf8_lossy(&escape_name_inner(name.as_bytes(), style, false))
-            .to_string()
-            .into()
-    }
 }
 
 /// Escape a directory name with respect to the given style.
 /// This is mainly meant to be used for ls' directory name printing and is not
 /// likely to be used elsewhere.
 pub fn escape_dir_name(dir_name: &OsStr, style: &QuotingStyle) -> OsString {
-    #[cfg(unix)]
+    let name = crate::os_str_as_bytes_lossy(dir_name);
-    {
+    crate::os_string_from_vec(escape_name_inner(&name, style, true))
-        let name = dir_name.as_bytes();
+        .expect("all byte sequences should be valid for platform, or already replaced in name")
-        OsStringExt::from_vec(escape_name_inner(name, style, true))
-    }
-    #[cfg(not(unix))]
-    {
-        let name = dir_name.to_string_lossy();
-        String::from_utf8_lossy(&escape_name_inner(name.as_bytes(), style, true))
-            .to_string()
-            .into()
-    }
 }
 
 impl fmt::Display for QuotingStyle {

src/uucore/src/lib/lib.rs

@@ -253,9 +253,10 @@ pub fn read_yes() -> bool {
     }
 }
 
-/// Helper function for processing delimiter values (which could be non UTF-8)
+/// Converts an `OsStr` to a UTF-8 `&[u8]`.
-/// It converts OsString to &[u8] for unix targets only
+///
-/// On non-unix (i.e. Windows) it will just return an error if delimiter value is not UTF-8
+/// This always succeeds on unix platforms,
+/// and fails on other platforms if the string can't be coerced to UTF-8.
 pub fn os_str_as_bytes(os_string: &OsStr) -> mods::error::UResult<&[u8]> {
     #[cfg(unix)]
     let bytes = os_string.as_bytes();
@@ -271,13 +272,28 @@ pub fn os_str_as_bytes(os_string: &OsStr) -> mods::error::UResult<&[u8]> {
     Ok(bytes)
 }
 
-/// Helper function for converting a slice of bytes into an &OsStr
+/// Performs a potentially lossy conversion from `OsStr` to UTF-8 bytes.
-/// or OsString in non-unix targets.
 ///
-/// It converts `&[u8]` to `Cow<OsStr>` for unix targets only.
+/// This is always lossless on unix platforms,
-/// On non-unix (i.e. Windows), the conversion goes through the String type
+/// and wraps [`OsStr::to_string_lossy`] on non-unix platforms.
-/// and thus undergo UTF-8 validation, making it fail if the stream contains
+pub fn os_str_as_bytes_lossy(os_string: &OsStr) -> Cow<[u8]> {
-/// non-UTF-8 characters.
+    #[cfg(unix)]
+    let bytes = Cow::from(os_string.as_bytes());
+
+    #[cfg(not(unix))]
+    let bytes = match os_string.to_string_lossy() {
+        Cow::Borrowed(slice) => Cow::from(slice.as_bytes()),
+        Cow::Owned(owned) => Cow::from(owned.into_bytes()),
+    };
+
+    bytes
+}
+
+/// Converts a `&[u8]` to an `&OsStr`,
+/// or parses it as UTF-8 into an [`OsString`] on non-unix platforms.
+///
+/// This always succeeds on unix platforms,
+/// and fails on other platforms if the bytes can't be parsed as UTF-8.
 pub fn os_str_from_bytes(bytes: &[u8]) -> mods::error::UResult<Cow<'_, OsStr>> {
     #[cfg(unix)]
     let os_str = Cow::Borrowed(OsStr::from_bytes(bytes));
@@ -289,9 +305,10 @@ pub fn os_str_from_bytes(bytes: &[u8]) -> mods::error::UResult<Cow<'_, OsStr>> {
     Ok(os_str)
 }
 
-/// Helper function for making an `OsString` from a byte field
+/// Converts a `Vec<u8>` into an `OsString`, parsing as UTF-8 on non-unix platforms.
-/// It converts `Vec<u8>` to `OsString` for unix targets only.
+///
-/// On non-unix (i.e. Windows) it may fail if the bytes are not valid UTF-8
+/// This always succeeds on unix platforms,
+/// and fails on other platforms if the bytes can't be parsed as UTF-8.
 pub fn os_string_from_vec(vec: Vec<u8>) -> mods::error::UResult<OsString> {
     #[cfg(unix)]
     let s = OsString::from_vec(vec);